Linear Predictive Clustering (LPC) partitions samples based on shared linear relationships between feature and target variables, with numerous applications including marketing, medicine, and education. Greedy optimization methods, commonly used for LPC, alternate between clustering and linear regression but lack global optimality. While effective for separable clusters, they struggle in non-separable settings where clusters overlap in feature space. In an alternative constrained optimization paradigm, Bertsimas and Shioda (2007) formulated LPC as a Mixed-Integer Program (MIP), ensuring global optimality regardless of separability but suffering from poor scalability. This work builds on the constrained optimization paradigm to introduce two novel approaches that improve the efficiency of global optimization for LPC. By leveraging key theoretical properties of separability, we derive near-optimal approximations with provable error bounds, significantly reducing the MIP formulation's complexity and improving scalability. Additionally, we can further approximate LPC as a Quadratic Pseudo-Boolean Optimization (QPBO) problem, achieving substantial computational improvements in some settings. Comparative analyses on synthetic and real-world datasets demonstrate that our methods consistently achieve near-optimal solutions with substantially lower regression errors than greedy optimization while exhibiting superior scalability over existing MIP formulations.

Since Multimodal Large Language Models (MLLMs) are increasingly being integrated into everyday tools and intelligent agents, growing concerns have arisen regarding their possible output of unsafe contents, ranging from toxic language and biased imagery to privacy violations and harmful misinformation. Current safety benchmarks remain highly limited in both modality coverage and performance evaluations, often neglecting the extensive landscape of content safety. In this work, we introduce OutSafe-Bench, the first most comprehensive content safety evaluation test suite designed for the multimodal era. OutSafe-Bench includes a large-scale dataset that spans four modalities, featuring over 18,000 bilingual (Chinese and English) text prompts, 4,500 images, 450 audio clips and 450 videos, all systematically annotated across nine critical content risk categories. In addition to the dataset, we introduce a Multidimensional Cross Risk Score (MCRS), a novel metric designed to model and assess overlapping and correlated content risks across different categories. To ensure fair and robust evaluation, we propose FairScore, an explainable automated multi-reviewer weighted aggregation framework. FairScore selects top-performing models as adaptive juries, thereby mitigating biases from single-model judgments and enhancing overall evaluation reliability. Our evaluation of nine state-of-the-art MLLMs reveals persistent and substantial safety vulnerabilities, underscoring the pressing need for robust safeguards in MLLMs.

Theoretically exploring the advantages of neural networks might be one of the most challenging problems in the AI era. An adaptive feature program has recently been proposed to analyze feature learning, the characteristic property of neural networks, in a more abstract way. Motivated by the celebrated Le Cam equivalence, we advocate the over-parameterized sequence models to further simplify the analysis of the training dynamics of adaptive feature program and present several pieces of supporting evidence for the adaptive feature program. More precisely, after having introduced the feature error measure (FEM) to characterize the quality of the learned feature, we show that the FEM is decreasing during the training process of several concrete adaptive feature models including linear regression, single/multiple index models, etc. We believe that this hints at the potential successes of the adaptive feature program.

The "Thinking with Text" and "Thinking with Images" paradigms significantly improve the reasoning abilities of large language models (LLMs) and Vision-Language Models (VLMs). However, these paradigms have inherent limitations. (1) Images capture only single moments and fail to represent dynamic processes or continuous changes, and (2) The separation of text and vision as distinct modalities, which hinders unified multimodal understanding and generation. Therefore, we propose "Thinking with Video", a new paradigm that leverages video generation models such as Sora-2 to use video frames as a unified medium for multimodal reasoning. To support this exploration, we developed the Video Thinking Benchmark (VideoThinkBench), which covers both vision-centric tasks (e.g., Eyeballing Puzzles) and text-centric tasks (e.g., GSM8K and MMMU). Our evaluation on VideoThinkBench establishes Sora-2 as a capable reasoner. On vision-centric tasks, Sora-2 is comparable to state-of-the-art (SOTA) VLMs, and even surpasses GPT-5 by 10% on eyeballing puzzles. On text-centric tasks, Sora-2 achieves 92% accuracy on MATH, and 69.2% accuracy on MMMU. Furthermore, we systematically analyze the source of these abilities. We also find that self-consistency and in-context learning can improve Sora-2's performance. In summary, our findings show that the video generation model is the potential unified multimodal understanding and generation model, positioning "Thinking with Video" as a potential unified multimodal reasoning paradigm.

Multimodal Large Language Models (MLLMs) struggle with continual learning, often suffering from catastrophic forgetting when adapting to sequential tasks. We introduce a routing-based architecture that integrates new capabilities while robustly preserving foundational knowledge. While Multi-Task Learning (MTL) offers a theoretical performance upper bound, it incurs a linearly scaling computational overhead as the number of tasks increases. In contrast, our method maintains fixed data and compute requirements regardless of the task sequence length. Across models ranging from 2B to 8B parameters, we demonstrate that our routing approach performs on par with MTL while retaining the training efficiency of sequential fine-tuning. Beyond merely mitigating forgetting, we observe that token-level routing facilitates cross-modal transfer, leveraging knowledge from one modality to bolster performance in another. Ablation studies confirm the approach's scalability: routing remains robust even with large expert pools and effectively capitalizes on task relatedness. Finally, we show that our method scales favorably, with larger models exhibiting minimal degradation compared to fully specialized fine-tuning.

The rapid evolution of AI-generated images poses growing challenges to information integrity and media authenticity. Existing detection approaches face limitations in robustness, interpretability, and generalization across diverse generative models, particularly when relying on a single source of visual evidence. We introduce AIFo (Agent-based Image Forensics), a training-free framework that formulates AI-generated image detection as a multi-stage forensic analysis process through multi-agent collaboration. The framework integrates a set of forensic tools, including reverse image search, metadata extraction, pre-trained classifiers, and vision-language model analysis, and resolves insufficient or conflicting evidence through a structured multi-agent debate mechanism. An optional memory-augmented module further enables the framework to incorporate information from historical cases. We evaluate AIFo on a benchmark of 6,000 images spanning controlled laboratory settings and challenging real-world scenarios, where it achieves 97.05% accuracy and consistently outperforms traditional classifiers and strong vision-language model baselines. These findings demonstrate the effectiveness of agent-based procedural reasoning for AI-generated image detection.

Whole-body humanoid motion represents a fundamental challenge in robotics, requiring balance, coordination, and adaptability to enable human-like behaviors. However, existing methods typically require multiple training samples per motion, rendering the collection of high-quality human motion datasets both labor-intensive and costly. To address this, we propose a data-efficient adaptation approach that learns a new humanoid motion from a single non-walking target sample together with auxiliary walking motions and a walking-trained base model. The core idea lies in leveraging order-preserving optimal transport to compute distances between walking and non-walking sequences, followed by interpolation along geodesics to generate new intermediate pose skeletons, which are then optimized for collision-free configurations and retargeted to the humanoid before integration into a simulated environment for policy adaptation via reinforcement learning. Experimental evaluations on the CMU MoCap dataset demonstrate that our method consistently outperforms baselines, achieving superior performance across metrics. Our code is available at: https://github.com/hhuang-code/One-shot-WBM.

Large Multimodal Models (LMMs) encode rich factual knowledge via cross-modal pre-training, yet their static representations struggle to maintain an accurate understanding of time-sensitive factual knowledge. Existing benchmarks remain constrained by static designs, inadequately evaluating LMMs' ability to understand time-sensitive knowledge. To address this gap, we propose MINED, a comprehensive benchmark that evaluates temporal awareness along 6 key dimensions and 11 challenging tasks: cognition, awareness, trustworthiness, understanding, reasoning, and robustness. MINED is constructed from Wikipedia by two professional annotators, containing 2,104 time-sensitive knowledge samples spanning six knowledge types. Evaluating 15 widely used LMMs on MINED shows that Gemini-2.5-Pro achieves the highest average CEM score of 63.07, while most open-source LMMs still lack time understanding ability. Meanwhile, LMMs perform best on organization knowledge, whereas their performance is weakest on sport. To address these challenges, we investigate the feasibility of updating time-sensitive knowledge in LMMs through knowledge editing methods and observe that LMMs can effectively update knowledge via knowledge editing methods in single editing scenarios.

Contact languages like English exhibit rich regional variations in the form of dialects, which are often used by dialect speakers interacting with generative models. However, can multimodal generative models effectively produce content given dialectal textual input? In this work, we study this question by constructing a new large-scale benchmark spanning six common English dialects. We work with dialect speakers to collect and verify over 4200 unique prompts and evaluate on 17 image and video generative models. Our automatic and human evaluation results show that current state-of-the-art multimodal generative models exhibit 32.26% to 48.17% performance degradation when a single dialect word is used in the prompt. Common mitigation methods such as fine-tuning and prompt rewriting can only improve dialect performance by small margins (< 7%), while potentially incurring significant performance degradation in Standard American English (SAE). To this end, we design a general encoder-based mitigation strategy for multimodal generative models. Our method teaches the model to recognize new dialect features while preserving SAE performance. Experiments on models such as Stable Diffusion 1.5 show that our method is able to simultaneously raise performance on five dialects to be on par with SAE (+34.4%), while incurring near zero cost to SAE performance.

Recent advances in Text-To-Speech (TTS) synthesis have achieved near-human speech quality in neutral speaking styles. However, most existing approaches either depend on costly emotion annotations or optimize surrogate objectives that fail to adequately capture perceptual emotional quality. As a result, the generated speech, while semantically accurate, often lacks expressive and emotionally rich characteristics. To address these limitations, we propose RLAIF-SPA, a novel framework that integrates Reinforcement Learning from AI Feedback (RLAIF) to directly optimize both emotional expressiveness and intelligibility without human supervision. Specifically, RLAIF-SPA incorporates Automatic Speech Recognition (ASR) to provide semantic accuracy feedback, while leveraging structured reward modeling to evaluate prosodic-emotional consistency. RLAIF-SPA enables more precise and nuanced control over expressive speech generation along four structured evaluation dimensions: Structure, Emotion, Speed, and Tone. Extensive experiments on Libri-Speech, MELD, and Mandarin ESD datasets demonstrate consistent gains across clean read speech, conversational dialogue, and emotional speech. On Libri-Speech, RLAIF-SPA consistently outperforms Chat-TTS, achieving a 26.1% reduction in word error rate, a 9.1% improvement in SIM-O, and over 10% gains in human subjective evaluations.

Inductive Knowledge Graph Reasoning (KGR) aims to discover facts in open-domain KGs containing unknown entities and relations, which poses a challenge for KGR models in comprehending uncertain KG components. Existing studies have proposed Knowledge Graph Foundation Models (KGFMs) that learn structural invariances across KGs to handle this uncertainty. Recently, Large Language Models (LLMs) have demonstrated strong capabilities for open-domain knowledge reasoning. As a result, the latest research has focused on LLM-based KGFMs that integrate LLM knowledge with KG context for inductive KGR. However, the intrinsic knowledge of LLMs may be overshadowed by sparse KG context, leading to LLM knowledge distortion, which can cause irreversible damage to model reasoning. Moreover, existing LLM-based KGR methods still struggle to fully constrain generative hallucinations in LLMs, severely limiting the credibility of reasoning results. To address these limitations, we propose a Knowledge Reasoning Language Model (KRLM) that achieves unified coordination between LLM knowledge and KG context throughout the KGR process. Specifically, we design a Knowledge Reasoning Language (KRL) instruction format and a KRL tokenizer to align LLM knowledge with KG representations. Then, we propose a KRL attention layer that coordinates intrinsic LLM knowledge with additional KG context through a dynamic knowledge memory mechanism. Finally, a structure-aware next-entity predictor is proposed, which strictly constrains the reasoning results within a trustworthy knowledge domain. Extensive experimental results on 25 real-world inductive KGR datasets demonstrate the significant superiority of the proposed KRLM\footnote{Our source codes are available at https://anonymous.4open.science/r/KRLM-EA36 in both zero-shot reasoning and fine-tuning scenarios.

Automating the formalization of mathematical statements for theorem proving remains a major challenge for Large Language Models (LLMs). LLMs struggle to identify and utilize the prerequisite mathematical knowledge and its corresponding formal representation in languages like Lean. Current retrieval-augmented autoformalization methods query external libraries using the informal statement directly, but overlook a fundamental limitation: informal statements lack direct mappings to mathematical theorems and lemmata, nor do those theorems translate trivially into the formal primitives of languages like Lean. To address this, we introduce DRIFT, a novel framework that enables LLMs to decompose informal mathematical statements into smaller, more tractable "sub-components". This facilitates targeted retrieval of premises from mathematical libraries such as Mathlib. Additionally, DRIFT retrieves illustrative theorems to help models use premises more effectively in formalization tasks. We evaluate DRIFT across diverse benchmarks (ProofNet, ConNF, and MiniF2F-test) and find that it consistently improves premise retrieval, nearly doubling the F1 score compared to the DPR baseline on ProofNet. Notably, DRIFT demonstrates strong performance on the out-of-distribution ConNF benchmark, with BEq+@10 improvements of 42.25% and 37.14% using GPT-4.1 and DeepSeek-V3.1, respectively. Our analysis shows that retrieval effectiveness in mathematical autoformalization depends heavily on model-specific knowledge boundaries, highlighting the need for adaptive retrieval strategies aligned with each model's capabilities.

Robust behaviour recognition in real-world farm environments remains challenging due to several data-related limitations, including the scarcity of well-annotated livestock video datasets and the substantial domain gap between large-scale pre-training corpora and agricultural surveillance footage. To address these challenges, we propose Cattle-CLIP, a domain-adaptive vision-language framework that reformulates cattle behaviour recognition as cross-modal semantic alignment rather than purely visual classification. Instead of directly fine-tuning visual backbones, Cattle-CLIP incorporates a temporal integration module to extend image-level contrastive pre-training to video-based behaviour understanding, enabling consistent semantic alignment across time. To mitigate the distribution shift between web-scale image-text data used for the pre-trained model and real-world cattle surveillance footage, we further introduce tailored augmentation strategies and specialised behaviour prompts. Furthermore, we construct CattleBehaviours6, a curated and behaviour-consistent video dataset comprising 1905 annotated clips across six indoor behaviours to support model training and evaluation. Beyond serving as a benchmark for our proposed method, the dataset provides a standardised ethogram definition, offering a practical resource for future research in livestock behaviour analysis. Cattle-CLIP is evaluated under both fully-supervised and few-shot learning scenarios, with a particular focus on data-scarce behaviour recognition, an important yet under-explored goal in livestock monitoring. Experiments show that Cattle-CLIP achieves 96.1% overall accuracy across six behaviours in supervised settings, with near-perfect recall for feeding, drinking and standing-ruminating behaviours, and demonstrates robust generalisation with limited data in few-shot scenarios.

Large language models (LLMs) exhibit strong symbolic and compositional reasoning, yet they struggle with time series question answering as the data is typically transformed into an LLM-compatible modality, e.g., serialized text, plotted images, or compressed time series embeddings. Such conversions impose representation bottlenecks, often require cross-modal alignment or finetuning, and can exacerbate hallucination and knowledge leakage. To address these limitations, we propose TS-Agent, an agentic, tool-grounded framework that uses LLMs strictly for iterative evidence-based reasoning, while delegating statistical and structural extraction to time series analytical tools operating on raw sequences. Our framework solves time series tasks through an evidence-driven agentic process: (1) it alternates between thinking, tool execution, and observation in a ReAct-style loop, (2) records intermediate results in an explicit evidence log and corrects the reasoning trace via a self-refinement critic, and (3) enforces a final answer-verification step to prevent hallucinations and leakage. Across four benchmarks spanning time series understanding and reasoning, TS-Agent matches or exceeds strong text-based, vision-based, and time-series language model baselines, with the largest gains on reasoning tasks where multimodal LLMs are prone to hallucination and knowledge leakage in zero-shot settings.

Multimodal encoders have pushed the boundaries of visual document retrieval, matching textual query tokens directly to image patches and achieving state-of-the-art performance on public benchmarks. Recent models relying on this paradigm have massively scaled the sizes of their query and document representations, presenting obstacles to deployment and scalability in real-world pipelines. Furthermore, purely vision-centric approaches may be constrained by the inherent modality gap still exhibited by modern vision-language models. In this work, we connect these challenges to the paradigm of hybrid retrieval, investigating whether a lightweight dense text retriever can enhance a stronger vision-centric model. Existing hybrid methods, which rely on coarse-grained fusion of ranks or scores, fail to exploit the rich interactions within each model's representation space. To address this, we introduce Guided Query Refinement (GQR), a novel test-time optimization method that refines a primary retriever's query embedding using guidance from a complementary retriever's scores. Through extensive experiments on visual document retrieval benchmarks, we demonstrate that GQR allows vision-centric models to match the performance of models with significantly larger representations, while being up to 14x faster and requiring 54x less memory. Our findings show that GQR effectively pushes the Pareto frontier for performance and efficiency in multimodal retrieval. We release our code at https://github.com/IBM/test-time-hybrid-retrieval

The tasks of idiom understanding and dialect understanding are both well-established benchmarks in natural language processing. In this paper, we propose combining them, and using regional idioms as a test of dialect understanding. Towards this end, we propose three new benchmark datasets for the Quebec dialect of French: QFrCoRE, which contains 4,633 instances of idiomatic phrases, and QFrCoRT, which comprises 171 regional instances of idiomatic words, and a new benchmark for French Metropolitan expressions, MFrCoE, which comprises 4,938 phrases. We explain how to construct these corpora, so that our methodology can be replicated for other dialects. Our experiments with 111 LLMs reveal a critical disparity in dialectal competence: while models perform well on French Metropolitan, 65.77% of them perform significantly worse on Quebec idioms, with only 9.0% favoring the regional dialect. These results confirm that our benchmarks are a reliable tool for quantifying the dialect gap and that prestige-language proficiency does not guarantee regional dialect understanding.

The rapid adoption of Large Language Models (LLMs) has raised significant environmental concerns. Unlike the one-time cost of training, LLM inference occurs continuously and dominates the AI energy footprint. Yet most sustainability studies report only coarse model-level metrics, treating energy efficiency as an afterthought rather than a primary objective. Addressing the limitation, we propose Component-Level Energy Assessment via Repetitions CLEAR, to overcome temporal mismatch between microsecond scale component execution and millisecond(ms) scale monitoring of energy sensors. Using CLEAR, we evaluate 15 models spanning four architecture types, keeping component-wise energy variance below 9.5% while capturing over 90% of total energy as individual components. We present the first comprehensive, fine-grained energy analysis of Transformer components across key parameters such as batch size, attention heads, hidden dimension, KV cache, and attention variants. Our findings reveal that Attention consumes significantly more Energy per FLOP as compared to the entire model, indicating that FLOPs alone fail to capture true component-level energy cost. CLEAR enables reliable fine-grained energy measurements and provides a strong formal foundation for predictive modelling of energy consumption.

Visually localizing an image, i.e., estimating its camera pose, requires building a scene representation that serves as a visual map. The representation we choose has direct consequences towards the practicability of our system. Even when starting from mapping images with known camera poses, state-of-the-art approaches still require hours of mapping time in the worst case, and several minutes in the best. This work raises the question whether we can achieve competitive accuracy much faster. We introduce FastForward, a method that creates a map representation and relocalizes a query image on-the-fly in a single feed-forward pass. At the core, we represent multiple mapping images as a collection of features anchored in 3D space. FastForward utilizes these mapping features to predict image-to-scene correspondences for the query image, enabling the estimation of its camera pose. We couple FastForward with image retrieval and achieve state-of-the-art accuracy when compared to other approaches with minimal map preparation time. Furthermore, FastForward demonstrates robust generalization to unseen domains, including challenging large-scale outdoor environments.

Although RLVR has become an essential component for developing advanced reasoning skills in language models, contemporary studies have documented training plateaus after thousands of optimization steps, i.e., notable decreases in performance gains despite increased computational investment. This limitation stems from the sparse exploration patterns inherent in current RLVR practices, where models rely on limited rollouts that often miss critical reasoning paths and fail to provide systematic coverage of the solution space. We present DeepSearch, a framework that integrates Monte Carlo Tree Search (MCTS) directly into RLVR training. In contrast to existing methods that rely on tree search only at inference, DeepSearch embeds structured search into the training loop, enabling systematic exploration and fine-grained credit assignment across reasoning steps. Through training-time exploration, DeepSearch addresses the fundamental bottleneck of insufficient exploration, which leads to diminishing performance gains over prolonged training. Our contributions include: (1) a global frontier selection strategy that prioritizes promising nodes across the search tree, (2) selection with entropy-based guidance that identifies confident paths for supervision, and (3) adaptive replay buffer training with solution caching for efficiency. Experiments on mathematical reasoning benchmarks show that DeepSearch achieves an average accuracy of 62.95\% and establishes a new state-of-the-art reasoning model, while using 5.7x fewer GPU hours than extended training approaches. These results highlight the importance of strategic exploration over brute-force scaling and demonstrate the promise of algorithmic innovation for advancing RLVR methodologies. DeepSearch establishes a new direction for scaling reasoning capabilities through systematic search rather than prolonged computation.

Dynamic resource allocation in open radio access network (O-RAN) heterogeneous networks (HetNets) presents a complex optimisation challenge under varying user loads. We propose a near-real-time RAN intelligent controller (Near-RT RIC) xApp utilising deep reinforcement learning (DRL) to jointly optimise transmit power, bandwidth slicing, and user scheduling. Leveraging real-world network topologies, we benchmark proximal policy optimisation (PPO) and twin delayed deep deterministic policy gradient (TD3) against standard heuristics. Our results demonstrate that the PPO-based xApp achieves a superior trade-off, reducing network energy consumption by up to 70% in dense scenarios and improving user fairness by more than 30% compared to throughput-greedy baselines. These findings validate the feasibility of centralised, energy-aware AI orchestration in future 6G architectures.

Reinforcement learning from human feedback (RLHF) has emerged as the standard paradigm for aligning large language models with human preferences. However, reward-based methods grounded in the Bradley-Terry assumption struggle to capture the nontransitivity and heterogeneity of real-world preferences. To address this, recent studies have reframed alignment as a two-player Nash game, giving rise to Nash learning from human feedback (NLHF). While this perspective has inspired algorithms such as INPO, ONPO, and EGPO that offer strong theoretical and empirical guarantees, they remain fundamentally restricted to two-player interactions, introducing a single-opponent bias that fails to capture the full complexity of realistic preference structures. This work introduces Multiplayer Nash Preference Optimization (MNPO), a novel framework that generalizes NLHF to the multiplayer regime. It formulates alignment as an n-player game, where each policy competes against a population of opponents while being regularized toward a reference model. We demonstrate that MNPO inherits the equilibrium guarantees of two-player methods while enabling richer competitive dynamics and improved coverage of diverse preference structures. Comprehensive empirical evaluation shows that MNPO consistently outperforms existing NLHF baselines on instruction-following benchmarks, achieving superior alignment quality under heterogeneous annotator conditions and mixed-policy evaluation scenarios. Together, these results establish MNPO as a principled and scalable framework for aligning LLMs with complex, non-transitive human preferences. Code is available at: https://github.com/smiles724/MNPO

Conventional deep neural nets (DNNs) initialize network parameters at random and then optimize each one via stochastic gradient descent (SGD), resulting in substantial risk of poor-performing local minima. Focusing on image interpolation and leveraging a recent theorem that maps a (pseudo-)linear interpolator Θ to a directed graph filter that is a solution to a corresponding MAP problem with a graph shift variation (GSV) prior, we first initialize a directed graph adjacency matrix A given a known interpolator Θ, establishing a baseline performance. Then, towards further gain, we learn perturbation matrices P and P(2) from data to augment A, whose restoration effects are implemented progressively via Douglas-Rachford (DR) iterations, which we unroll into a lightweight and interpretable neural net. Experiments on different image interpolation scenarios demonstrate state-of-the-art performance, while drastically reducing network parameters and inference complexity.

Assessing the reliability of Large Language Models (LLMs) by confidence elicitation is a prominent approach to AI safety in high-stakes applications, such as healthcare and finance. Existing methods either require expensive computational overhead or suffer from poor calibration, making them impractical and unreliable for real-world deployment. In this work, we propose GrACE, a Generative Approach to Confidence Elicitation that enables scalable and reliable confidence elicitation for LLMs. GrACE adopts a novel mechanism in which the model expresses confidence by the similarity between the last hidden state and the embedding of a special token appended to the vocabulary, in real-time. We fine-tune the model for calibrating the confidence with targets associated with accuracy. Extensive experiments show that the confidence produced by GrACE achieves the best discriminative capacity and calibration on open-ended generation tasks without resorting to additional sampling or an auxiliary model. Moreover, we propose two confidence-based strategies for test-time scaling with GrACE, which not only improve the accuracy of the final decision but also significantly reduce the number of required samples, highlighting its potential as a practical solution for deploying LLMs with reliable, on-the-fly confidence estimation.

Assistants on assembly tasks show great potential to benefit humans ranging from helping with everyday tasks to interacting in industrial settings. However, evaluation resources in assembly activities are underexplored. To foster system development, we propose a new multimodal QA evaluation dataset on assembly activities. Our dataset, ProMQA-Assembly, consists of 646 QA pairs that require multimodal understanding of human activity videos and their instruction manuals in an online-style manner. For cost effectiveness in the data creation, we adopt a semi-automated QA annotation approach, where LLMs generate candidate QA pairs and humans verify them. We further improve QA generation by integrating fine-grained action labels to diversify question types. Additionally, we create 81 instruction task graphs for our target assembly tasks. These newly created task graphs are used in our benchmarking experiment, as well as in facilitating the human verification process. With our dataset, we benchmark models, including competitive proprietary multimodal models. We find that ProMQA-Assembly contains challenging multimodal questions, where reasoning models showcase promising results. We believe our new evaluation dataset contributes to the further development of procedural-activity assistants.

Recent advances in interactive video generations have demonstrated diffusion model's potential as world models by capturing complex physical dynamics and interactive behaviors. However, existing interactive world models depend on bidirectional attention and lengthy inference steps, severely limiting real-time performance. Consequently, they are hard to simulate real-world dynamics, where outcomes must update instantaneously based on historical context and current actions. To address this, we present Matrix-Game 2.0, an interactive world model generates long videos on-the-fly via few-step auto-regressive diffusion. Our framework consists of three key components: (1) A scalable data production pipeline for Unreal Engine and GTA5 environments to effectively produce massive amounts (about 1200 hours) of video data with diverse interaction annotations; (2) An action injection module that enables frame-level mouse and keyboard inputs as interactive conditions; (3) A few-step distillation based on the casual architecture for real-time and streaming video generation. Matrix Game 2.0 can generate high-quality minute-level videos across diverse scenes at an ultra-fast speed of 25 FPS. We open-source our model weights and codebase to advance research in interactive world modeling.

Facial representation pre-training is crucial for tasks like facial recognition, expression analysis, and virtual reality. However, existing methods face three key challenges: (1) failing to capture distinct facial features and fine-grained semantics, (2) ignoring the spatial structure inherent to facial anatomy, and (3) inefficiently utilizing limited labeled data. To overcome these, we introduce PaCo-FR, an unsupervised framework that combines masked image modeling with patch-pixel alignment. Our approach integrates three innovative components: (1) a structured masking strategy that preserves spatial coherence by aligning with semantically meaningful facial regions, (2) a novel patch-based codebook that enhances feature discrimination with multiple candidate tokens, and (3) spatial consistency constraints that preserve geometric relationships between facial components. PaCo-FR achieves state-of-the-art performance across several facial analysis tasks with just 2 million unlabeled images for pre-training. Our method demonstrates significant improvements, particularly in scenarios with varying poses, occlusions, and lighting conditions. We believe this work advances facial representation learning and offers a scalable, efficient solution that reduces reliance on expensive annotated datasets, driving more effective facial analysis systems.

Vision Language Models (VLMs) encode multimodal inputs over large, complex, and difficult-to-interpret architectures, which limit transparency and trust. We propose a Multimodal Inversion for Model Interpretation and Conceptualization (MIMIC) framework that inverts the internal encodings of VLMs. MIMIC uses a joint VLM-based inversion and a feature alignment objective to account for VLM's autoregressive processing. It additionally includes a triplet of regularizers for spatial alignment, natural image smoothness, and semantic realism. We evaluate MIMIC both quantitatively and qualitatively by inverting visual concepts across a range of free-form VLM outputs of varying length. Reported results include both standard visual quality metrics and semantic text-based metrics. To the best of our knowledge, this is the first model inversion approach addressing visual interpretations of VLM concepts.

Tasks that require character-level reasoning, such as counting or locating characters within words, remain challenging for contemporary language models. A common conjecture is that language models' reliance on subword units, rather than characters, contributes to their struggles with character-level tasks, yet recent studies offer conflicting conclusions about the role of tokenization, leaving its impact unclear. To address this gap, we introduce CharBench, a comprehensive benchmark of character-level tasks that is two orders of magnitude larger than existing alternatives. We evaluate a diverse range of leading open-weight and proprietary models on CharBench and find that it presents a significant challenge to modern LLMs, with an average accuracy of 43.6% and 32.3% on some tasks. We present an in-depth analysis of how intrinsic properties of words and their segmentations into tokens correspond to model performance. For counting tasks, we find that tokenization properties are weakly correlated with correctness, while the length of the queried word and the actual character count play a more significant part. In contrast, for tasks requiring intra-word positional understanding, performance is negatively correlated with the length of the token containing the queried character, suggesting that longer tokens obscure character position information for LLMs. We encourage future work to build on the benchmark and evaluation methodology introduced here as tools for improving model performance on such tasks.

Quantifying aleatoric uncertainty in medical image segmentation is critical since it is a reflection of the natural variability observed among expert annotators. A conventional approach is to model the segmentation distribution using the generative model, but current methods limit the expression ability of generative models. While current diffusion-based approaches have demonstrated impressive performance in approximating the data distribution, their inherent stochastic sampling process and inability to model exact densities limit their effectiveness in accurately capturing uncertainty. In contrast, our proposed method leverages conditional flow matching, a simulation-free flow-based generative model that learns an exact density, to produce highly accurate segmentation results. By guiding the flow model on the input image and sampling multiple data points, our approach synthesizes segmentation samples whose pixel-wise variance reliably reflects the underlying data distribution. This sampling strategy captures uncertainties in regions with ambiguous boundaries, offering robust quantification that mirrors inter-annotator differences. Experimental results demonstrate that our method not only achieves competitive segmentation accuracy but also generates uncertainty maps that provide deeper insights into the reliability of the segmentation outcomes. The code for this paper is freely available at https://github.com/huynhspm/Data-Uncertainty

Large Language Models (LLMs) are widely used to generate plausible text on online platforms, without revealing the generation process. As users increasingly encounter such black-box outputs, detecting hallucinations has become a critical challenge. To address this challenge, we focus on developing a hallucination detection framework for black-box generators. Motivated by the observation that hallucinations, once introduced, tend to persist, we sample future contexts. The sampled future contexts provide valuable clues for hallucination detection and can be effectively integrated with various sampling-based methods. We extensively demonstrate performance improvements across multiple methods using our proposed sampling approach.